How do I light up LED's according to audio frequency bands?

[Dave Pollum]

I'd like to have LED's light according to the tones/frequency in
music. Sort of a light show type of thing, I guess. I'm a digital
guy, but I guess I'd start with active filters to divide music (~100Hz
to 15 kHz) into 8, 10, 12 bands ? Then the output from each band
would be converted into PWM to drive some multi-colored LED's.
I know this description is vague and my idea is half-baked, so I'd be
grateful to hear about some real ideas
TIA !
-Dave Pollum

 

[David L. Jones]

I'd like to have LED's light according to the tones/frequency in
music. Sort of a light show type of thing, I guess. I'm a digital
guy, but I guess I'd start with active filters to divide music (~100Hz
to 15 kHz) into 8, 10, 12 bands ? Then the output from each band
would be converted into PWM to drive some multi-colored LED's.
I know this description is vague and my idea is half-baked, so I'd be
grateful to hear about some real ideas
TIA !
-Dave Pollum

The way you have described is the way it's been mostly done in the
distant past.
Disadvantage is that you need a bandpass filter for every channel you
want, that adds up to a lot of circuitry.
These days it's done with FFT (Fast Fourier Transform) processing,
everything is done in software. A DSP (Digital Signal Processor) is
often used, but normal 8/16bit microcontrollers like PIC and Atmel
et.al have enough grunt to do it these days.

Dave.

 

[MooseFET]

I'd like to have LED's light according to the tones/frequency in
music.  Sort of a light show type of thing, I guess.  I'm a digital
guy, but I guess I'd start with active filters to divide music (~100Hz
to 15 kHz) into 8, 10, 12 bands ?  Then the output from each band
would be converted into PWM to drive some multi-colored LED's.
I know this description is vague and my idea is half-baked, so I'd be
grateful to hear about some real ideas

It works better if you filter and rectify each band and then process
the rectified values a bit before you light the LEDs

Each band's results should have a little of the nearby bands results
subtracted from it. This narrows the bandwidth of the effect of each
filter.

The rectified results are filtered in a filter that averages the value
over many cycles but also subtracts away a bit of the value from the
history. This makes the LEDs respond to changes a bit more than they
otherwise would.

 

[Eeyore]

I'd like to have LED's light according to the tones/frequency in
music. Sort of a light show type of thing, I guess. I'm a digital
guy, but I guess I'd start with active filters to divide music (~100Hz
to 15 kHz) into 8, 10, 12 bands ? Then the output from each band
would be converted into PWM to drive some multi-colored LED's.
I know this description is vague and my idea is half-baked, so I'd be
grateful to hear about some real ideas
TIA !

Yes, that would work.

Graham

 

[Eeyore]

The way you have described is the way it's been mostly done in the
distant past.

It's how I built one nearly 40 years ago, except the filter outputs drove
triacs which phase angle modulated mains voltage lighting. Only 3 bands
though.

Don't forget the isolation transformer on the audio in !

Graham

 

[Dave Pollum]

The way you have described is the way it's been mostly done in the
distant past.
Disadvantage is that you need a bandpass filter for every channel you
want, that adds up to a lot of circuitry.
These days it's done with FFT (Fast Fourier Transform) processing,
everything is done in software. A DSP (Digital Signal Processor) is
often used, but normal 8/16bit microcontrollers like PIC and Atmel
et.al have enough grunt to do it these days.

Dave.

Hmmm... lots to think about. I'll have to research FFT and DSP and
take another look at the YouTube videos for ideas.
Thanks, all =)
-Dave Pollum

 

[neon]

LEDS are diodes arsenic or whatever material they are made from. there is a minimumn voltage required to turn them on and also a minimumn current. There is power dissipation of the devices involved. exceed that and blew it.
since they are diodes they follows an exponential curve just like a diode when it gets .7v volts more current by a factor of 10 will not make much difference on the .7v. And difference of collor means difference of voltage drop to be considered a red 1.8v a blue 3-4v the brightnes cannot very well be controlled as opposed to incandense bulbs. the range of brightness is not enough to get the results that you need. But try it you might like it. cheaper xmas lights with optics and triacs direct from AC. Furthermore a led or lamps flashing at 10-15kc is like it is on all the time.

 

[[email protected]]

I'd like to have LED's light according to the tones/frequency in
music.  Sort of a light show type of thing, I guess.  I'm a digital
guy, but I guess I'd start with active filters to divide music (~100Hz
to 15 kHz) into 8, 10, 12 bands ?  Then the output from each band
would be converted into PWM to drive some multi-colored LED's.
I know this description is vague and my idea is half-baked, so I'd be
grateful to hear about some real ideas
TIA !
-Dave Pollum

Sure Dave, this is easy, just intermodulate the audio with a sweep
signal, this is how old style spectrum analyzers worked. It's not to
hard to hook up.

 

[MooseFET]

Sure Dave, this is easy, just intermodulate the audio with a sweep
signal, this is how old style spectrum analyzers worked. It's not to
hard to hook up.

This idea does not work for audio bands because the sweep rate must be
too slow to be useful.

I just had an interesting idea for someone digital.

In music, harmonics are heard as the character of the tone not as
individual tones. The brain picks out the low frequency component as
the underlying note.

Here what I suggest just for fun:

Take the audio input and convert it into a delta modulated signal at
lets say 5MHz.

Get a "top octave generator" and use it to make square waves on the
octave below middle C.

Use XORs to combine the delta mod signal with each of the notes of the
octave.

The next step as analog to make the digital easier to see:
The XORed signals can then be hooked to simple RC low pass filters to
"delta demodulate" them. This then goes to a comparitor

The next step as digital:
An up down counter can be used to make a simple IIR filter like
circuit as follows.

An up down counter controlled directly by the output of the XOR would
be an integration. If the system is clocked at a multiple of the 5MHz
that is used in the delta modulator, we can use some of the extra
clocks to make the counter tend to settle back towards zero.

To make the settling process work, but not very like the real RC,
simply use the MSB of the counter as a sign value and one out of every
so many clock cycles direct the sign to the up/down control.

Another XOR between the upper two bits of the counters would control
the LEDs.

 

[Vladimir Vassilevsky]

MooseFET wrote:

Here what I suggest just for fun:

Take the audio input and convert it into a delta modulated signal at
lets say 5MHz.

Get a "top octave generator" and use it to make square waves on the
octave below middle C.

Use XORs to combine the delta mod signal with each of the notes of the
octave.

[...]

Very elegant indeed.
Why don't we see you in comp.dsp?

VLV

 

[GregS]

It's how I built one nearly 40 years ago, except the filter outputs drove
triacs which phase angle modulated mains voltage lighting. Only 3 bands
though.

Don't forget the isolation transformer on the audio in !

Graham

The OP did not explain much. My first color organ I made from a tube radio
receiver out of a Northrop RCAT drone. I was in that division. Others had tried,
but I made it work. I think ony 4 channels. Its not just technology, but
its mostly artwork, getting a good display using the proper colors, and getting the right
bandpass. Using more midrange frequiencies makes piano and guitar work more
interesting. I later was trying to use the Op-Amp bandpass method, but
I never finished that project, using #47 lamps. In the disco years I saw interesting
displays in Mexico. I have never seen pinwheels after that. I made a pinwheel, or wagon wheel
driver system, and the beat of the music controled the rotation direction, and whether
the pinwheel was expanding or contracting.

greg

 

[MooseFET]

Here what I suggest just for fun:

Take the audio input and convert it into a delta modulated signal at
lets say 5MHz.

Get a "top octave generator" and use it to make square waves on the
octave below middle C.

Use XORs to combine the delta mod signal with each of the notes of the
octave.

[...]

Very elegant indeed.
Why don't we see you in comp.dsp?

I have a wife and a job and a garage to put shelves up in and I sleep
8 hours a day.

 

[MooseFET]

Here what I suggest just for fun:

Take the audio input and convert it into a delta modulated signal at
lets say 5MHz.

Get a "top octave generator" and use it to make square waves on the
octave below middle C.

Use XORs to combine the delta mod signal with each of the notes of the
octave.

[...]

Very elegant indeed.
Why don't we see you in comp.dsp?

BTW: There is a mistake. We need a second XOR running at 90
degrees. The 90 degree signal can be made by XORing F with 2*F so the
top octave circuit can be run at a higher clock and divided down.

Many years back I worked on the design of a sonar where I suggested
the same sort of method.